איך יכולים לשפר תיאורי היעדים ללמידת ריפוד מותנית?

תקציר מקורי באנגליתarXiv:2609.39901v1 Announce Type: new Abstract: Goal-conditioned reinforcement learning (GCRL) relies heavily on how target goals are represented to the policy. While recent methods encode goals via temporal distance, occupancy, or controllability, it remains unclear how much downstream performance actually depends on representation quality. We study this in offline GCRL by constructing an exact temporal-distance goal representation in deterministic mazes. We then systematically corrupt its geometric quality while keeping the downstream learner fixed. Across OGBench navigation tasks and two algorithms, large changes in goal-representation quality produce almost no change in performance. However, applying the same interventions to the agent's current state more than doubles success, reveali